[Nolug] wireless mesh networks, MAN clouds and user overload

Saturday at Dot's we discussed the idea of wireless clouds, where
theoretically one could route from one end of town to the other via some sort
of mesh network (http://sensorsmag.com/articles/0203/38/main.shtml). I
mentioned that there exists at least one Linux distro that acts in a
"predatory" manner and works reasonably well
(http://www.locustworld.com/modules.php?op=modload&name=Downloads&file=index).

Someone made the comment that too many users would kill throughput. I don't
really know how the mesh networks are engineered, but it occurs to me that
part of the spec or implementation could set aside bandwidth and other
resources to compensate for a glut of users. Some sort of prioritizing, QoS,
whatever.

Caching of data (store-and-forward) might work for some traffic, but I'm not
sure that store-and-forward can even work at all; at any rate, this would
take a lot of hands-on administration, which might put it out of the running
altogether.

Also mentioned was latency due to large numbers of hops. This would become a
real problem when routing VoIP, for example.

Another issue, which isn't totally clear in my mind, had to do with routing
from one AP (for lack of a better word) to another, and a decrease of space
in each packet for actual data. Judson was saying something about this, but I
think I was away from the table at that point. Probably it had to do with
bridging instead of routing, but perhaps he can share with the group exactly
what he was talking about, if he has the time to post that here.

-- 
Joey Kelly
< Minister of the Gospel | Computer Networking Consultant >
http://joeykelly.net
___________________
Nolug mailing list
nolug@nolug.org